Fine-tuning, multi-agent, and DPO approaches all improve cognitive and linguistic authenticity of LLM-simulated math students over few-shot prompts, with each offering distinct benefits for teacher noticing of student thinking as revealed in interviews.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.HC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Developing Authentic Simulated Learners for Mathematics Teacher Learning: Insights from Three Approaches with Large Language Models
Fine-tuning, multi-agent, and DPO approaches all improve cognitive and linguistic authenticity of LLM-simulated math students over few-shot prompts, with each offering distinct benefits for teacher noticing of student thinking as revealed in interviews.